SHOULDERMEN had been in my wishlist for probably six months or more when I noticed it was on sale and I finally decided to see what it was all about. The promise of an AI-powered game intrigued me, largely because this is the promised-land frontier of new, character-driven gameplay that may offer the next big innovation to our beloved medium. Now, while SHOULDERMEN doesn't necessarily promise all that in its pitch, it does have a unique one regardless: convince both an angel and a devil, via voice or by text, to say a password that will free them from an otherworldly prison (which is haunted by a wonderfully cheerful, intentionally nagging repetitive jingle). Being able to have some kind of cohesive, unscripted back and forth with two characters is something that a lot of people would like to do in a game. Even if just a simple exercise, that's worth trying out - right? Well, having spent what I feel is enough time with SHOULDERMEN, I'm not so sure.
It's a simple but surprisingly messy experience. The characters frequently contort to face the moving camera angles, awkwardly drifting in and out of frame with no consistency. I frequently found myself typing my questions to them, as the voice recognition bugs out just as frequently and fails to pick up on sentences (and I'm using a Shure SM7B with an interface/filter, no background noise on my end). But that didn't stop me accidentally opening the console by pressing the apostrophe key, apologies for my insistence on good grammar. That's not even to mention the surprisingly hefty performance toll the game has due to its Unreal Engine basis, the uninspired asset usage and the general lack of personality to the aesthetics of SHOULDERMEN.
Is this latter point a fair critique? Well, maybe it's best described as a tech demo, rather than a "game". There's a couple of reasons for that. Firstly - it's so simple and one-dimensional in your mission and interaction with the characters that it's hard to feel this is a skill-based endeavour that requires that level of investment from the player. Sure, SHOULDERMEN doesn't have all that much going for it, but beyond a woman with a halo, dressed in gold and a man in a suit, with horns, there's no real way to relate to these characters as anything more than the AI models they actually are. It's not so much a puzzle game as it is a conversation with an LLM, thinly veiled with bought-asset aesthetics. But would changing said aesthetics affect that much here? I don't really think so. At first, you score an easy win with the angel character, easily coaxing them into stating your password. But then you're left with the "HARD" mode character: the devil, who dances around the password with glee (admittedly, it's quite funny!). Even in times when I purposefully addressed him first, rather than the angel character, I found myself more bored and infuriated than I was driven to get him to relent. And that's not because there's a lack of progression or content here, with a reloading of the one level SHOULDERMEN has offering the same fundamental experience whether you win or give up (but with a new password). Nor was it down to an expectation of ease - I think it's good that there's a considerable challenge here, contrasting the promise of success (even without a reward).
Ultimately, I think the problem that SHOULDERMEN has is that it's just not really fun. I managed to "trick" the devil character twice into saying the password, but both times I got there, I didn't feel like it happened from the encouragement of a unique gameplay system, or from my own ingenuity. Rather, I felt like I got lucky with the draw - I banged my head against the wall and then a sudden dent appeared. The second time was particularly silly, repeating random words from my prior questions, only for him to repeat them back to me with the password attached. No rhyme or reason, just a seemingly random confession. Therein lies the biggest flaw: unpredictability. Rather than a sense of gradual progress, like you're at least getting slightly more familiar with the characters and the methodology of coaxing them towards a particular statement, as a consequence of asking questions, the greater feeling is one of somehow finding a light switch in the dark. Yes, the lights are on, but you didn't expect it, so now you're temporarily blinded. Good, basic games can work with player skill - Donkey Kong's cabinet relied on your ability to time jumps over barrels, climb ladders and user power-ups to your advantage. You knew what was coming, so you could deal with it as and when, growing more adept with practice. If there is such a system for player growth built into SHOULDERMEN, it's not at all apparent in a way that the game's tutorial clues about "context and intuition" meaningfully hint towards as much as they vaguely shrug at the suggestion - you'll figure it out, maybe.
It's a shame because, truthfully, my expectations were not that high for a title that cost me £0.47. But I think there's much more nuance and finesse needed to really make this game feel the way it should, to draw players back into its repetitive and simple structure and invite them into a challenge that feels fun, rather than infuriating, to engage with.
